Using remotely sensed solar radiation data for reference evapotranspiration estimation at a daily time step

Abstract

Solar radiation is an important climatic variable for assessing reference evapotranspiration ( E 0), but it is seldom available in weather station records. Meteosat satellite images processed with the Heliosat-2 method provide the HelioClim-1 database, which displays spatialized solar radiation data at a daily time step for Europe and Africa. The aim of the present work was to investigate the interest of satellite-sensed solar radiation for E 0 calculation, where air temperature is the sole local weather data available. There were two study areas in Southern France. One (Southwest, SW) is characterized by Oceanic climate and the other (Southeast, SE) by Mediterranean climate. A data set of daily values for 19 weather stations spanning five years (2000–2004) was used. First, a sensitivity analysis of the Penman–Monteith formula to climate input variables was performed, using the Sobol’ method. It shows that E 0 is mainly governed by solar radiation during summer, and by wind speed during winter. Uncertainties of HelioClim-1 solar radiation data and their repercussions on E 0 formulae were evaluated, using the FAO-56 Penman–Monteith formulae (PM) and radiation-based methods (Turc, TU; Priestley–Taylor, PT and Hargreaves-Radiation, HR). It was shown that HelioClim-1 data slightly underestimate solar radiation and provide relative RMSE (root mean square error) of 20% of the mean annual value for SW and 14% for SE. The propagation of HelioClim-1 data uncertainties is small in PM but considerable in radiation methods. Four estimation methods were then compared to PM data: the 1985 Hargreaves formula (HT) based on air temperature only; TU, PT and HR, based on air temperature and satellite-sensed solar radiation. Radiation methods were more precise and more accurate than HT, with RMSE ranging from 0.52 mm to 0.86 mm against 0.67–0.96 mm. These results suggest that using satellite-sensed solar radiation may improve E 0 estimates for areas where air temperature is the only available record at ground level.